THE CLOUD DYNAMICS AND RADIATION DATABASE A FOCUS ON OROGRAPHIC PRECIPITATION

Transcription

1 THE CLOUD DYNAMICS AND RADIATION DATABASE A FOCUS ON OROGRAPHIC PRECIPITATION by Joseph A. Hoch A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science (Atmospheric and Oceanic Sciences) at the UNIVERSITY OF WISCONSIN-MADISON 2006

2 i ABSTRACT Passive microwave remote sensing of precipitation from platforms such as the Special Sensor Microwave Imager (SSM/I), the Advanced Microwave Scanning Radiometer (AMSR), and the Tropical Rainfall Measurement Mission (TRMM) has been a major focus in hydrological research for the past several years. Estimation of precipitation from these platforms relies on the accuracy of the particular retrieval algorithm being utilized. Retrieval algorithms are based on cloud radiation databases (CRDs) that relate in-situ measurements of brightness temperatures to a-priori microphysical profiles located in CRDs. A potential problem with a CRD retrieval based approach is that profiles can be chosen that are unrepresentative of the dynamical and thermodynamical state of the atmosphere. Recently, the concept of the Cloud Dynamics and Radiation Database (CDRD) for precipitation retrieval purposes has been introduced. The CDRD is an improved version of current CRDs. The CDRD contains the same information as present CRDs, but in addition contains information about the dynamical and thermodynamical structure of the atmosphere. The CDRD contains dynamical and thermodynamical tags that are computed from a cloud-resolving model. The cloud resolving model used to build the CDRD is the University of Wisconsin Non-Hydrostatic Modeling System (UW-NMS). These simulations are also used to calculate brightness temperatures and microphysical profiles. During a particular retrieval global forecast models can be used to more accurately mine

3 ii the CDRD database, retrieve a more accurate microphysical profile, and thus improve precipitation retrieval. The main objective of this thesis is to first discuss the methodology for implementing the CDRD system over the entire globe and to show the possibility of retrieving useful subsets of information from a large global database system. The primary technique for extracting useful subsets of information is through the use of data mining techniques. Several data mining techniques are presented and discussed. Orographically enhanced precipitation is fairly challenging to accurately represent from microwave platforms and the current CRD approach. A detailed case study over central and southern California focuses on orographically enhanced precipitation retrieval. The CDRD leads to more accurate microphysical profiles retrieved for the particular event. Finally, Bayes Theorem, along with the CDRD, is used to more accurately predict the probability of snowfall over the western United States for a particular storm. The CDRD is a robust system that improves microwave precipitation retrieval techniques. This system can also be used for many other earth science applications. In the future the CDRD will be used to investigate the relationships between microphysical quantities and atmospheric parameters.

4 iii ACKNOWLEDGEMENTS I would like to thank all the people involved in helping me accomplish the maximum amount possible over the past two years. During my time here, I have increased my theoretical knowledge of meteorology and also my love for the weather. First, I would like to thank my advisor, Dr. Greg Tripoli. I have been able to learn a lot from interacting with him over the past two years. He provided just the right amount of guidance for my research, while also letting me working independently. Another benefit of working with Greg was that I could join the AOS bowling team. Whenever the weather was not that exciting, Greg and I could always talk bowling strategy. I would also like to thank my research project colleagues, Dr. Carlo Medaglia, Dr. Amita Mehta, Dr. Alberto Mugnai, and Dr. Eric Smith. It has been a pleasure working with them and I hope to continue collaborations in the future. I am definitely looking forward to visiting Carlo in Rome, Italy. Thanks also to all my friends I have made over the past two years for all of your support. Anytime work would get a little crazy they were always there to get away from it all. Also, thank you to all the professors I have had the opportunity to interact with over the past two years, there are far too many to list by name. My choice to attend the University of Wisconsin turned out to be a great decision. I will never forget the football games, the tornado chasing, the hurricane chasing, the skiing, and the opportunity to grow as a scholar in the field of meteorology.

5 iv TABLE OF CONTENTS 1. Introduction 1 2. Precipitation Retrieval Mechanisms 5 a. Microwave Satellite Platforms 5 b. The Goddard Profiling Algorithm 8 3. Cloud Dynamics and Radiation Database (CDRD) CDRD Modeling Systems 16 a. Bayes Theorem 16 b. University of Wisconsin Non-Hydrostatic Modeling System 17 c. Successive Order of Interaction (SOI) Radiative Transfer Model Orographic Precipitation 20 a. Previous Field Studies 24 th 6. Case Study - January 7-11 th, 2005 California Storm 27 a. Upper-Levels The Jet Stream 29 b. Mid-Levels Large Scale Potential Vorticity 33 c. Low-Levels Moisture and 1km Winds 36 d. Precipitation Verification 39 e. Sensitivity Experiments CDRD Data Mining CDRD Retrieval Applications Conclusions References 68

6 1 CHAPTER 1: INTRODUCTION Passive microwave remote sensing of precipitation from space is a relatively new concept in the field of meteorology. Not until the late 1970 s and early 1980 s were the first attempts made to accurately retrieve rainfall estimates from space. Two of the earliest microwave platforms were Nimbus-6 and Nimbus-7. After the launch of these satellites, research in precipitation retrieval from space flourished (Wilheit 1976, Liou 1979, Prabhakara 1986, Wilke 1986, Spencer 1987, Petty 1990, 1992). Since the Nimbus program many other microwave satellite platforms have been launched, including the Scanning Multichannel Microwave Radiometer [SMMR, ] (Prabhakara 1986), the Special Sensor Microwave/Imager [SSM/I, 1987 present] (Ferriday 1994), the Advanced Microwave Scanning Radiometer [AMSR, 2002 present] (McCollum 2005), and the Tropical Rainfall Measurement Mission Microwave Imager [TRMM, 1997 present] (Kummerow et al. 1998). Improvement of passive microwave remote sensing of precipitation is still a key scientific goal. A new National Aeronautics Space Agency (NASA) mission is currently planned that will further enhance the capabilities of precipitation remote sensing (i.e. the Global Precipitation Measurement Mission [GPM]). Current microwave remote sensing platforms return radiance measurements over an observational satellite footprint. These radiance measurements can be converted to corresponding brightness temperatures. The most common range for microwave frequencies on these platforms is from 10 GHz through 89 GHz. In current retrieval schemes, brightness temperatures are matched to corresponding precipitation rates with similar microphysical precipitation structure. Precipitation retrieval schemes require Cloud Resolving Model (CRM) simulations of several different types of precipitation

7 2 systems, precipitation structure, and the microphysical properties of the simulated environment. One of the most commonly used microwave precipitation retrieval algorithms is the Goddard Profiling Algorithm [GPROF] (Kummerow 2001). Retrieval schemes make use of a-priori databases that relate microwave frequency brightness temperatures to surface precipitation rates through simulated microphysical specifications (hydrometeor sizes, shapes, and distributions). These databases are produced from CRM simulations. GPROF utilizes a cloud radiation database (CRD) that was produced from the Goddard Cumulus Ensemble Model (GCE) and the University of Wisconsin Non-Hydrostatic Modeling System (UW-NMS). The CRD contains vertical estimates of the simulated microphysical structure of the atmosphere, commonly referred to as microphysical profiles. Estimates of the six categories of hydrometeors (i.e. rain, snow, pristine ice, graupel, aggregates, and cloud) are contained within each profile. Although shown to be fairly accurate, passive microwave remote sensing retrieval schemes could be improved. Often various configurations of hydrometeors can produce similar brightness temperatures. Thus, microphysical properties taken from the a-priori CRD can be mixed from differing atmospheric environments. Orographic precipitation is one particularly challenging area for current retrieval schemes. This problem could be caused by the CRD database consisting of mainly tropical simulations (Kummerow 2001). The small scale variability and shallow precipitation structure could also cause some of the biggest retrieval errors. Satellites often look straight through the shallow liquid water path of orographic storms. The overall goal of this research is to develop a new retrieval scheme to improve the accuracy of matching observed brightness temperatures and a-priori microphysical

8 3 precipitation structure. The Cloud Dynamics and Radiation Database (CDRD) is the proposed improvement on the current CRD approach. The CDRD includes all of the typical information of a CRD, but also simulated dynamical and thermodynamical tags. At the time of retrieval, microphysical profiles can be selected from both observed satellite measurements and dynamical/ thermodynamical tags to estimate surface precipitation rates. By using a tag-based retrieval approach, constrained by a Bayesian algorithm, there is potential to reduce the variability of retrieved microphysical profiles, thus improving precipitation retrieval. Improving satellite precipitation retrieval is essential for accurate measurement and understanding of the global water cycle. More accurate retrieval can also be used for data assimilation purposes and better model forecasts. The proposed CDRD techniques improve the accuracy and usefulness of retrieval from current satellite platforms. This thesis is organized as follows. Chapter 2 discusses currently available passive microwave remote sensing platforms. These instruments include SSM/I, AMSR, and TRMM. The upcoming Global Precipitation Measurement mission is also discussed. An overview of GPROF is presented, along with the necessity for the CDRD approach. Chapter 3 explains the CDRD concept in more detail. Microphysical profile, dynamic tag, and thermodynamic tag variables are presented. Database tags are selected based on their ability to distinguish differing atmospheric environments. The associated CRM nested 3-grid setup is discussed. Chapter 4 describes the predictive models used to formulate the CDRD system. These models include the University of Wisconsin Non-Hydrostatic Modeling System and the Successive Order of Interaction (SOI) radiative transfer model. The setup for

9 4 these models is discussed, along with various other details. Also, Bayes theorem is presented as a useful technique to search through the CDRD database system. Chapter 5 focuses on the theory of orographically enhanced precipitation. As mentioned, orographic precipitation is a challenging quantity to accurately diagnose from satellite measurements. Past research and orographic field studies are presented to determine appropriate CDRD tags to correctly diagnose orographically enhanced precipitation. Chapter 6 presents an orographic precipitation case study. The UW-NMS is used to simulate a severe orographic event that occurred from January 7 th 11 th, 2005 over the Sierra Nevada region of California. The case study includes several sensitivity tests, along with examples of data input into the CDRD. Retrieval of CDRD profiles is accomplished through data mining techniques. Data mining is a powerful tool used to search through large database systems, commonly referred to as the data warehouse. Chapter 7 presents commonly used mining techniques and the current CDRD approach. Data mining is one of the main highlights of the CDRD retrieval approach because of its effectiveness and efficiency. Chapter 8 presents an overview of potential CDRD applications. A simulated retrieval of microphysical profiles over the California case study region is discussed. Using the CDRD approach reduces the amount of variance in retrieved microphysical profiles, thus leading to more accurate profiles. Also, a global application using the CDRD for snowfall probability retrieval is discussed. Through the use of CDRD tags and Bayes Theorem, the level of certainty in precipitation retrieval can be increased. Finally, conclusions and future work are discussed.

10 5 CHAPTER 2: PRECIPITAITON RETRIEVAL ALGORITHMS Currently passive microwave remote sensing platforms, available for precipitation retrieval, are AMSR, SSM/I, and TRMM. In upcoming years, a follow-up mission to TRMM will be launched, the Global Precipitation Measurement mission. Potential problems with the retrieval algorithms and the need for a CDRD approach are discussed. a. Microwave Satellite Platforms The Special Sensor Microwave/Imager, launched in 1987, provides an opportunity to study over 15 years of data. There have been six different SSM/I platforms, all carefully inter-calibrated. The SSM/I is part of the Defense Meteorological Satellite Program (DMSP) satellite program. Data from the SSM/I is often used to calculate ocean wind speeds, water vapor fields, cloud water fields, and surface rain rates. The SSM/I is part of the NASA Pathfinder program. The resolution of the 85.5 GHz channel is 12.5km. The satellites are in sun-synchronous, circular, and nearly polar orbits. The swath width is approximately 1400km, with global coverage. The Advanced Microwave Scanning Radiometer was launched on May 4 th, 2002, on the NASA s Aqua spacecraft. Over the oceans, AMSR can measure several different parameters such as sea surface temperatures (SST), wind speeds, water vapor, cloud water, and rain rates. Mean spatial resolutions decrease with increasing center frequencies from 56km to 5.4km for the 89.0 GHz channel. This represents a significant increase in resolution from the 12.5km resolution of the SSM/I. The satellite footprint of

11 6 AMSR for the 89.0 GHz channel is 6 x 4 km. This is a large increase compared to the 15 x 13km footprint for SSM/I. The Tropical Rainfall Measurement Mission was launched in November The purpose of this mission was to map precipitation variability over the entire tropical belt. The TRMM not only contains a microwave imager, but also an onboard radiometer. The TRMM Microwave Imager (TMI) is very similar to the SSM/I radiometer. The TMI measures the intensity of radiation at five separate frequencies, 10.7, 19.4, 21.3, 37, and 85.5 GHz, dual polarization. These frequencies are similar to those of the SSM/I, except that TMI has the additional 10.7 GHz channel designed to provide a more-linear response towards high rainfall rates. Another improvement of TMI is the increased resolution. This is mainly a function of a lower orbit altitude. TMI has a 487 mile (780-kilometer) wide swath on the surface. The higher resolution of TMI, as well as the additional 10.7 GHz frequency, makes TMI a better instrument than its predecessors. The TRMM K U Precipitation Radar (PR) measures reflectivity over a narrow swath. Reflectivity measurements can be used to infer rain rates via a Z-R relationship. The PR direct precipitation measurements are then used to calibrate the wide measurement swath of the TMI. TRMM has been a vital mission used to produce frequency distributions of rainfall intensity and coverage, partition stratiform and convective precipitation events, measure the vertical distribution of hydrometeors, and make tropical measurements of latent heat release. TRMM has also aided in the study of atmospheric teleconnections, such as the Madden-Julian oscillation, the Asian monsoon, and ENSO events.

12 7 TRMM has provided invaluable data over the tropical regions of the globe. The need now exists to study precipitation outside of the tropical regions. To accomplish this goal, NASA is planning the next generation satellite mission, the Global Precipitation Measurement mission. The Global Precipitation Measurement mission (GPM) will continue the work of TRMM. This satellite is an improved version of TRMM, including increased spatial resolution and increased coverage area. One of the biggest improvements is the addition of dual-frequency precipitation radar (DPR). The DPR will have a 5km resolution at 13.6 GHz K U band and 35.5 GHz K A band. The GPM radars will have the ability to measure, via reflectivity and estimates of attenuation, the vertical profiles of clouds and precipitation, including the drop size distribution. The GPM Microwave Imager (GMI) will have a swath width of 850km. It contains the same frequencies as TRMM at higher spatial resolutions and four additional high frequency millimeter-wave channels at about 166 GHz and 183 GHz. The current scheduled launch date of GPM is Space-borne microwave technology has significantly improved from SSM/I through TRMM. Resolutions and spatial coverage have both increased. Table 1 shows a summary of the available microwave frequencies on these instruments. These frequencies are used by precipitation algorithms to estimate surface precipitation rates. In formulating the CDRD system these channels are all considered, with the majority included.

13 8 SSM/I AMSR TRMM GPM 19.4 V&H V&H 10.7 V&H 10.7 V&H 22.2 V&H V&H 19.4 V&H 19.0 V&H 37.0 V&H 18.7 V&H 21.3 H 22.0 V&H 85.5 V&H 23.8 V&H 37.0 V&H 37.0 V&H 36.5 V&H 85.5 V&H 85.0 V&H 50.3 V 150 V&H 89.0 V&H V&H Table 1. Available microwave frequencies on current satellites. b. The Goddard Profiling Algorithm (GPROF) The GPROF rain rate algorithm uses statistical inversion techniques, based upon theoretically calculated relations between rainfall rates and brightness temperatures. Inverse hydrometeor profiles are used to explicitly account for the potential errors introduced in the GPROF theoretical calculations. This is accomplished by allowing various vertical distributions in the theoretical brightness temperature calculations and by requiring consistency between the observed and calculated brightness temperatures (Kummerow and Giglio 1994). Smith et al. (1998) states that the cloud model generated profiles in the GPROF algorithm are assigned a-priori probabilities. These are related to brightness temperatures at all SSM/I frequencies and polarizations through a forward RTE model. The GPROF rain rate algorithm explicitly accounts for the vertical structure of precipitation within a cloud. This vertical structure is a key factor in determining the upwelling radiances. The algorithm simulates the upwelling radiances by means of a radiative transfer scheme (Kummerow and Giglio 1994).

14 9 Different configurations of hydrometeors can produce similar brightness temperatures. Thus, a potential problem with the CRD retrieval is that particular profiles, featuring the best match to observed brightness temperatures, can be taken from differing atmospheric conditions. Microphysical profiles could be combined from stratiform and convective events, cold frontal and tropical systems, unstable and stable atmospheres, or other potential situations. If inaccurate matching occurs, the microphysical profile produced from the CRD weighting functions would be an inaccurate representation of the true microphysical structure of the atmosphere, which leads to errors in the rainfall estimation. The proposed solution is to develop an improved version of current CRDs. The CDRD, discussed in more detail in the next chapter, includes all the typical information of a CRD, but also dynamical and thermodynamical tags paired with each microphysical profile. At the time of retrieval, profiles are obtained not only from satellite brightness temperatures but also a wide variety of tags. These tags should reduce the amount of unrepresentative environmental profiles selected.

15 10 CHAPTER 3: CLOUD DYNAMICS AND RADIATION DATABASE (CDRD) An operational global database has been implemented using daily randomly selected simulations beginning on August 1 st, A global database, including randomly selected simulations, is necessary for several reasons. First, it is important to obtain a robust database warehouse with many possible meteorological conditions. In order to utilize this system for global precipitation retrieval, all meteorological events such as stratiform, convective, tropical, frontal, cellular, orographic, etc. must be represented in the CDRD. Second, using daily simulations captures seasonal variations. Finally, random CRM simulations, using the UW-NMS, are important to eliminate any biases towards particular precipitation events or geographic regions. A two-way nested grid structure is used with three grids for CDRD simulations. The innermost domain size of these simulations is approximately 5 degrees by 5 degrees. The resolutions of these grids are 50km, 10km, and 2km respectively. When a CRM simulation begins, a 12Z Global Forecasting System (GFS), 9-hour model precipitation prediction is used to sample the selected random innermost domain. This prediction is necessary to test if the selected location has the potential to produce precipitation in the mesoscale simulation. This GFS test is mainly used because of computational resources. Current mesoscale simulations take approximately 24 hours to produce a 12 hour forecast. If a region shows no sign of precipitation in the GFS forecast a new random location is selected. The UW-NMS performs a 12 hour simulation over the selected location. Microphysical profiles, dynamical, and thermodynamical tags for the CDRD database are saved at the 12 hour simulation time. Information is stored in the database only for the

16 11 12-hour forecast time to allow for spin-up and local forcing to develop. Often it takes 12 hours for accurate tropical cyclone spin-up. Microphysical profiles are saved based on simulated surface precipitation rates. The criteria for saving a profile is taken to be when surface rain rates are 0.50 mm hr -1 or greater and/or a frozen (snow, graupel, aggregates, pristine crystals) surface precipitation rates are 0.25 mm hr -1 or greater. These criteria are selected based on the capability of current microwave remote precipitation sensors. These precipitation criteria are near the accepted lower limits of useful satellite data (Wilheit 2003). The following table shows the variables that make up a microphysical profile. The corresponding data is saved at all 36 vertical model levels at each appropriate grid point. As discussed earlier, these microphysical profiles are necessary to convert radiance measurements to precipitation estimates. Hydrometeor Measurements (Rain, Snow, Graupel, Aggregate, Pristine Crystals) Mixing Ratio (g/kg) Concentration (#/cm 3 ) Diameters (micrometer) Terminal Velocity (cm/s) Densities (g/cm 3 ) Surface Rate (mm/hr) Other Profile Variables Total Condensate Mixing Ratio (g/kg) Water Vapor Mixing Ratio (g/kg) Cloud Water Mixing Ratio (g/kg) Surface Skin Temperature (K) Latent Heating Term (K/day) Diabatic Moisture Term (K/day) Pressure (hpa) Height (m) Temperature (K) Zonal Wind (m/s) Meridional Wind (m/s) Vertical Velocity (m/s) Liquid Coating Flag Table 2. UW-NMS variables included in a standard microphysical profile.

17 12 Dynamical and thermodynamical tags are paired with microphysical profiles at two different resolutions, 50km grid spacing (low resolution) and 2km grid spacing (high resolution). The majority of variables are saved at low resolution so that they are comparable to global model resolutions, such as the Global Forecasting System (GFS). New precipitation retrievals will use microwave brightness temperatures paired with the following CDRD dynamical tags, obtainable from global operational models, which are shown in the next two tables. These tags should provide for more accurate microphysical profiles selected from the CRM database, thus improving precipitation estimation. The selection of CDRD profiles is based on a Bayesian approach. Bayesian techniques, discussed further in a later chapter, use the dynamical tags to more accurately sample available profiles.

18 13 50KM GRID SPACING TAGS Mean Sea Level Pressure (hpa) Surface Temperature (F) Freezing Level (m) Lifted Index (K) Surface Theta Gradient (K/m) 700mb Theta Gradient (K/m) LFC Height (m) LCL Height (m) **U-Wind (m/s) Froude Number Surface Theta-E Topography Gradient (K/m) Height (m) **V-Wind (m/s) U Momentum Flux (kg/ms 2 ) Surface Theta-E (K) Surface Brunt Vaisala Frequency (s -1 ) 700mb Theta-E Gradient (K/m) ** Q Vector Convergence PBL Height (m) Richardson Number in the PBL V Momentum **Temperature Surface 2 Flux (kg/ms ) (K) -1 Divergence (s ) CIN (J/kg) Maximum CAPE (J/kg) Surface CAPE (J/kg) Kinetic Energy (J) Potential Vorticity at 700 and 200 mb (PVU) Surface Vertical Vorticity (s -1 ) Vertical Vorticity at 700 and 200 mb (s -1 ) 0-6km Wind Shear (s -1 ) Divergence at 700 and 200 mb (s -1 ) **Vertical Velocity (m/s) Theta-E minimum (K) 500 and 850 mb thickness (m) Potential Vorticity Advection at 700 and 250 mb (Kkgm 3 /s 2 ) Height of Maximum CAPE (m) Diabatic Moisture Term (K/day) Latent Heat Term (K/day) **Specific Humidity (g/kg) Table 3. Large Scale (50km) Dynamic/Thermodynamic Tags included in the CDRD.v1 system. ** Represents vectors (1000, 925,850,700,500,250,200,150,100 mb)

19 14 5KM GRID SPACING TAGS Cloud Ceiling (m) Topography Height (m) Largest Topography Neighbor Difference (m) Direction of Topography Direction (degrees) PBL Height (m) Mean Sea Level Pressure (hpa) Surface Pressure (hpa) ** Temperature (K) ** Q Vector Convergence (m 2 /skg) ** Vertical Velocity (m/s) Cloud Fraction Convective Cloud Fraction Stratiform Cloud Fraction Table 4. Small Scale (5km) Dynamic/Thermodynamic Tags included in the CDRD.v1 system. ** Represents vectors (1000, 925,850,700,500,250,200,150,100 mb) The dynamical/ thermodynamical tags were selected based on their ability to differentiate between particular atmospheric environments. For example, consider the following meteorological events and the appropriate CDRD tags: Convective vs. Stratiform Precipitation Vertical Velocity, Cloud Fraction, Wind Shear, CAPE Tropical or Non-Tropical Environment MSLP, Temperature, Theta-E, Freezing Level, Thickness, Latitude Severe Weather Environment CAPE, CIN, Lifted Index, Theta-E, Vertical Velocity, Wind Shear, Divergence Inertial Gravity Wave Environment Momentum Flux, Temperature, Froude Number, Brunt-Vaisala Frequency Mid-latitude Storm MSLP, Temperature, Specific Humidity, Winds, Potential Vorticity, Vertical Velocity, Thickness

20 15 Orographic Influence Topography Height, Topography Slope, Winds, Momentum Flux QG-Forced Motion Q-Vector Convergence, Divergence, Potential Vorticity Advection, Temperature Microwave frequencies included in the database are 6.6 GHz, GHz, 18.7 GHz, GHz, 23.8 GHz, 36.5 GHz, 85.5 GHz, 89.0 GHz, 150 GHz, and GHz. All of these channels are dual polarization, vertical and horizontal, except 23.8 GHz. These channels were selected based on present and future satellite platforms, as discussed in the previous chapter.

21 16 CHAPTER 4: CDRD MODELING SYSTEMS The following section outlines three different types of models used in the CDRD design. The first model discussed is a mathematical model, Bayes theorem. This mathematical model is the basis for future CDRD precipitation retrieval algorithms. GPROF also uses Bayesian statistical techniques to obtain the best estimated microphysical profiles. The other two models are atmospheric models, the UW-NMS and the SOI radiative transfer model. a. Bayes Theory The CDRD is designed to support a Bayesian framework for precipitation retrieval. A Bayesian approach is similar to a statistical inversion algorithm that achieves a maximum likelihood estimate, while being trained by the CDRD. Bayes theorem is used to obtain certain optimal parameters, such as microphysical profiles, from a set of measurements, i.e. the dynamical tags. Bayes theorem, for the purpose of rain rate retrievals, can be expressed as the following: P ( R Tb) = P( R) P( Tb R) Equation 1. Bayes Theorem expressed in terms of rain rate (R) and brightness temperatures (TB). Vertical distributions of hydrometeor profiles are expressed by R and vertical distribution of brightness temperatures are expressed by Tb. The goal of the CDRD retrieval scheme is to find a particular hydrometeor profile given a certain brightness temperature, P(R Tb). The probability that a certain hydrometeor profile will be observed, P(R), is

22 17 produced from the mesoscale model. The second term on the right side of the equation, P(Tb R) is produced from radiative transfer schemes. The benefits of a Bayesian based approach are shown in a later chapter with an operational CDRD retrieval. b. University of Wisconsin Non-Hydrostatic Modeling System The UW-NMS is used to create mesoscale simulations for the CDRD input. The model is described in full detail by Tripoli (1992). One of the most fundamental reasons for using the UW-NMS is its unique terrain following system, variable step topography. The following figure is an example of this terrain system. Figure 1. Three different mesoscale modeling terrain systems. The UW-NMS uses a unique Variable Step Condition to specify terrain. Other mesoscale models represent terrain in various ways such as terrain following systems or step topography systems. Terrain following systems do not accurately resolve

23 18 very steep topography, as shown in figure 1. Step topography systems have difficulty resolving subtle changes in topography, as shown in the same figure. The UW-NMS terrain following system is capable of handling both these scenarios very well. Since the UW-NMS invokes such an advanced terrain system, it is capable of reproducing various terrain induced flows. Tripoli shows many classic terrain problems modeled accurately by the UW-NMS (1992). In a later chapter the UW-NMS is used to study an orographically enhanced precipitation event over California. The UW-NMS uses a two-way nesting scheme. The largest grid is usually taken to be hydrostatic, unless non-hydrostatic information exists. The outer grid initial data may be interpolated from another model run such as the NCEP or European Center for Medium range Weather Forecast (ECMWF) models, or may be initialized from a horizontally homogeneous state, such as a sounding (Tripoli 1992). All CDRD simulations are initialized from Global Forecasting System (GFS) analysis files. The large-scale dynamic tags are taken from the outer grid while the high resolution tags are from the inner grid domain.

24 19 c. Successive Order of Interaction (SOI) Radiative Transfer Model The amount of time required running model simulations globally and building a complete database requires a fast, yet fairly accurate, radiative transfer code for the purposes of this project. The Successive Order of Interaction (SOI) Radiative Transfer Model, developed by Ralf Bennartz at the University of Wisconsin, is utilized for computation of microwave brightness temperatures. This model computes a brightness temperature field for a given frequency and polarization in about two minutes. The SOI is a 1-dimensional azimuthally-averaged, plane-parallel radiative transfer model. This model includes the effects of scattering from all hydrometeors. The SOI ignores atmospheric polarization, but not surface polarization (Heidinger et al. 2004). This is a hybrid model which uses both the doubling method and the Neumann method, which involves successive order of scattering, to obtain the upwelling radiance. The figure below illustrates the structure of the SOI model. Notice the doubling and successive order scattering properties of the model. Figure 2. Doubling method used in the Successive Order of Interaction (SOI) radiative transfer model. CHAPTER 5: OROGRAPHIC PRECIPITATION

25 20 CHAPTER 5: OROGRAPHIC PRECIPITATION Orographic precipitation is the cause of many extreme flooding events over mountainous regions worldwide and a challenge to accurately measure from space-based instruments. This chapter highlights the mechanisms of orographic precipitation formation and the potential dynamic and thermodynamic tags that should be included in the CDRD system. Some previous orographic field experiments are discussed to determine the appropriate CDRD tags for detection of this type of precipitation. In this section the CDRD tags are highlighted in bold. CDRD tags can be used to diagnose all the known types of orographic precipitation. These tags, along with satellite measurements, can identify previously undetectable precipitation. Seven mechanisms for the production of orographic precipitation, documented by Houze (1993), are now discussed. Figure 3 is an example of these seven different orographic precipitation mechanisms. Figure 3. Seven possible mechanisms leading to the formation of orographic precipitation.

26 21 The first mechanism is the Seeder-Feeder. During this processes ice crystals, from a glaciated stratiform cloud aloft, fall onto a cloud of supercooled droplets, which were initially generated by the orography. The glaciation process is referred to as the Bergeron-Findeisen process. Water droplets are most prevalent (at temperatures warmer than about -20 o C) but ice crystals are more efficient centers of growth from the vapor stage. This follows from the difference between the values of saturation vapor pressure over an ice surface when compared with a water surface. In mixed clouds, containing both liquid and ice, the air is close to being saturated with respect to liquid water, but is super-saturated (an unstable phase) with respect to ice. Consequently, in mixed clouds, ice crystals grow from the vapor phase much more rapidly than do the nearby droplets. This growth method, first documented by Bergeron (1935), is the Bergeron - Findeisen process. Vertical temperature structure is a key variable to recognize this process and is included in the CDRD. The second orographic mechanism involves the presence of stably stratified air. This air can be forced to rise over the mountain range. As the air rises, it cools, becomes saturated, and precipitation can result. This is one of the most common types of orographic rainfall, especially in the mid-latitudes (Houze 1993). The condition of whether a given air parcel will go over a mountain is given by the Froude number (Fr), as shown in equation 3. Fr = U h N m U = Wind Speed h m = Mountain Height N = Brunt-Väisälä frequency Equation 2. Froude Number calculation.

27 22 In most simple terms, the Froude number can be thought of as a ratio of kinetic energy to potential energy. If the Froude number is greater than 1 the air parcel will make it over the mountain, if it is less than 1 then it will not, and if it is equal to 1 then the air parcel reaches the mountain top with zero velocity. For low Froude number flows (less than 1) this simple physical reasoning suggests that the flow is essentially blocked by the topography and must either go around or be turned back. For this precipitation formation mechanism it is assumed that the Froude number is greater than one. The Froude number and topography height are included in the CDRD to diagnose this type of orographic influence. The third mechanism to form orographic precipitation involves a potentially unstable atmosphere. For potential instability theta-e must decrease with height, as shown in equation 4. dθ e dz < 0 Equation 3. Criteria that defines a potential unstable atmosphere. The mountain acts as a trigger to the convection. As air is forced to rise in this unstable environment, deep convection can occur. Often thunderstorms are found between the foothills and the crest of the mountain range. Theta-E and Theta-E gradients are included in the CDRD system. The fourth precipitation formation mechanism involves the creation of a surface boundary, such as in temperature, dew point, and/or wind. The mountain can create a

28 23 pseudo dryline effect and trigger deep convection downstream. A sharp gradient in temperature or dew point can cause a gradient of mass from each air mass. The lighter air mass is forced to rise over the more dense air. This rising air often then triggers convection. A gradient in wind speeds may produce enough shear to supplement long lived thunderstorms. Temperature, dew point, wind, and wind shear are all dynamic variables included in the CDRD. The fifth mechanism is very similar to the fourth described above. In the presence of calm large-scale winds, especially at low levels, the mountain may induce boundarylayer flows. These flows develop due to the difference in daytime heating and nighttime cooling. Convergence of these flows may trigger thunderstorm formation. Houze (1993) notes that this is the leading mechanism for orographic precipitation enhancement near the Inter-Tropical Convergence Zone (ITCZ). The final two mechanisms, leeside convection enhancement and convection triggering, are often overlooked mechanisms for the formation of orographic rainfall. These mechanisms are especially prevalent near an isolated mountain peak. If air is stable and can not be forced over the mountain, which implies the Froude number is less than one, it must go around the mountain. As air is forced around the mountain leeside convergence may result in updrafts and/or rainfall enhancement. Convection can be triggered due to this convergence. Again, this mechanism highlights the importance of including the Froude number in the CDRD.

29 24 These seven mechanisms are thought to be responsible for the formation of orographic precipitation. In the following case study the second orographic precipitation mechanism is responsible for enhancement of precipitation. a. Previous field studies With a process as complex as orographic precipitation, the best way to gain understanding of the phenomenon is to take real-time measurements and analyze the data collected. Numerous field programs have focused on trying to better understand the process of orographic precipitation. Three such field programs are now discussed in more detail to understand which dynamical tags should be included in the CDRD. The Mesoscale Alpine Programme Intensive Observing Period 2B (MAP IOP-2B) took place from September 18 th - 21 st, 1999, the Intermountain Precipitation Experiment (IPEX) took place from January 31 st th - February 25, 2000, and the California Landfalling Jets Experiment (CALJET) took place during the winter of 1997/98. A heavy precipitation event, forced by orography, occurred on September 19 th and 20 th, 1999 which was observed and later modeled by Lin et al. (2004). This event occurred over the Alps, near the Lago Maggiore region, during the MAP IOP-2B period. The Pennsylvania State University/ National Center for Atmospheric Research Mesoscale Modeling system (PSU/NCAR MM5) is used as a modeling tool for this event. During this event, a deep trough helped to evaporate moisture from the Adriatic Sea and advect moisture into the Alps region. One of the key results from this study is the importance of a strong impinging low level jet (LLJ), coupled with the influx of a moisture source. Lin

30 25 et al. (2004) showed that the low-level convergence created by the mountain barrier creates a convergence zone which helps increase precipitation amounts. Shafer et al. (2005) looked at the influence terrain has on synoptic and mesoscale precipitation distributions during IPEX. One of the goals of this field experiment was to advance the understanding about orographically enhanced precipitation, with a focus on the Wasatch Mountains of Utah. This field experiment focused on data-assimilation for the purpose of improving mesoscale model predictions of quantitative precipitation forecasts over complex terrain. Shafer et al. (2005) studied an occluded front approaching the Sierra Nevada mountain range. In this study the occluded front could not surmount the mountain range, while the upper-level trough moved over the mountain region unimpeded. This resulted in a discontinuous low-level storm evolution. This study is important because it shows that the low-level structure of a midlatitude cyclone can be strongly influenced by terrain, while upper-level features move relatively unimpeded by the terrain. The final field project and most applicable to this research is the California Landfalling Jets Experiment, which occurred during the winter of Neiman et al. (2002) studies the relationship between upslope flow and rainfall in the California coastal mountain region. Neiman et al. (2002) looks at individual cases, low-level jet cases and the entire winter season. Linear correlations between average upslope flow components versus rain rate are as high as 0.94 for individual storm cases. The low-level jet cases are around 0.75 and the entire winter season is around 0.75 (Neiman et al. 2002). This research study shows that the layer of flow that best influences orographic precipitation is near the mountain, around 1 km above sea level for the California coast. Neiman et al.

31 26 (2002) discusses that orographic precipitation is also controlled by ambient thermodynamic stratification and dynamic flow around topography, availability of moisture, precipitation formation efficiency, and latent heat release. On the simplest level, orographic precipitation is controlled by the upslope flow influence (Neiman et al. 2002). From these field experiments variables such as available moisture, low-level winds, vertical pressure distribution, terrain slope, theta-e, and latent heat release are shown to be important for detecting various types of orographic precipitation. All of these variables are included in the CDRD system. The following chapter shows the ability of the UW-NMS to predict orographic events and an application of the CDRD for a simulated orographic event.

32 27 CHAPTER 6: CASE STUDY - CALIFORNIA OROGRAPHIC STORM The period of January 7 th through January 11 th, 2005 brought tremendous amounts of rainfall and snowfall throughout much of California. Areas such as Nordhoff Ridge and Opids Camp recorded over 25 inches of equivalent rainfall during the event. The storm caused millions of dollars in damage and at least 10 deaths due to mudslides. This storm was particularly damaging for the region as a similar storm system had just impacted the area from December 26 th through January 5 th, The main branch of the jet stream brought a low-pressure system off the coast of British Columbia. A relatively strong subtropical jet stream provided the necessary moisture over the entire region. The high-pressure system further west in the Pacific Ocean set up a blocking pattern that kept the jet stream in the relatively same position for several days. Another low-pressure system became stationary off the coast of California, helping to further stream subtropical moisture into the region. During the four day period the precipitable water reached values as high as 1.5 inches. As impulses moved onto the California coast winds flowing counterclockwise intersected the Sierra Nevada mountain chain and caused heavy orographic precipitation enhancement. As is the case with many storms in California, the mountainous terrain of the region enhanced the precipitation. Previous studies have commonly focused on the Sierra Nevada region because of the regular occurrence of this type of precipitation enhancement (Marwitz 1987; Reynolds and Kuciauskas 1988; Colle 2004; Dettinger et al. 2004; Galewsky and Sobel 2005). The following case study is used to highlight the application of the CDRD system and the ability of the UW-NMS mesoscale model to simulate orographic precipitation.

33 28 The UW-NMS is used to simulate the four-day event. A three grid nesting system is used with an outer grid resolution of 50km, middle grid resolution of 10km, and an inner grid resolution of 2km. The largest time step on the outermost grid is 90 seconds, 22.5 seconds on the middle grid nest, and seconds on the innermost grid. The UW- NMS is particularly good at simulating orographically enhanced precipitation due to its variable step topography system. First, the upper-level dynamics of the storm are analyzed. The focus is on the polar and subtropical jet streams and their impacts during the event. Second, the midlevel dynamics are analyzed using potential vorticity as a diagnostic tool for approaching short waves. Finally, the low-level dynamics are analyzed using 1-km wind flow and surface moisture over the region. All of these variables are included as CDRD tags. Sensitivity experiments are used to diagnose the amount of orographic influence during the event. It is necessary to determine the sensitivity of the parameters in question to determine if they should be included in the CDRD.

34 29 a. Upper-Levels The Jet Stream The jet stream plays an important role in most large scale synoptic weather patterns. Jet streams exist due to the baroclinicity of the atmosphere, such as the temperature gradient between the equator and North Pole. An examination of the thermal wind equation, shown in the following equation, can explain the existence of such a feature. Equation 4. Thermal wind equation. V T Thermal Wind f Coriolis Parameter φ Geopotential Height The thermal wind circles areas of cold air, for example the North Pole. Thus, in the northern hemisphere winds are westerly. The thermal wind equation shows that these westerly winds will increase with height, thus the existence of the jet stream. This region of fast moving winds is found usually around 12 13km. The most common jet stream is the polar jet. However, a second jet can exist due to the advection of momentum from the equator northwards. This jet stream is commonly referred to as the sub-tropical jet stream. The sub-tropical jet stream is usually found at latitudes of 30 to 40 degrees in general westerly flow, and is important for the transport of moisture. Jet streams are important for several reasons. First, they influence the direction of synoptic scale features. Second, they can create regions of upper level divergence and convergence. Thirdly, they can influence moisture transport. Finally, they can influence stability and be responsible for gravity wave formation.

35 30 Both the polar and sub-tropical jet streams play an important role in the California orographic storm development. Shown in figure 4 is the 3-dimensional wind speed field, as predicted by the UW-NMS. The daily progression of the two jet streams, beginning on January 7 th at 12Z through January 11 th at 12Z, can be seen. Also shown in figure 4 is a schematic of the large-scale synoptic setup for this event. This image is courtesy of the California Nevada River Forecast Center. The UW-NMS simulation is consistent with this depiction. The polar jet stream digs a deep trough off the Pacific coast. The jet stream is rather strong from the 7 th until the 10 th with wind speeds in excess of 75 m/s. The polar jet stream begins to weaken and move northward by the 10 th th and 11, as shown in figure 4. The jet stream was basically stationary for 4 straight days continuously steering synoptic disturbances towards the California coast. As mentioned previously, the sub-tropical jet is also important in this case. The steady sub-tropical jet increases the moisture transport from the equator towards the California coast. As shown in figure 4, the sub-tropical jet stream is rather strong as well nearing values of 60 m/s. Typically, the sub-tropical jet is much weaker than the polar jet. During this event the two jet streams seem to couple, especially on January 8 th and 9 th. This coupling leads to enhanced jet stream dynamics over the region. During the same two days large-scale gravity waves, caused by the jet stream interaction with the Sierra Nevada and Rocky mountain chains, seem to be evident. Finally, the associated upper-level divergence caused by the jet stream is an important consideration in this case. Figure 4 shows the favored regions of upper-level divergence caused by the sub-tropical jet. This region is commonly referred to as the left

36 31 exit region. Upper level divergence can lead to enhanced vertical motion due to the conservation of mass.

37 32 Polar Sub Tropical Jan 7th Gravity Waves Jan 8th Jan 9th Jan 10th Jan11th Figure 4. Daily progression of the polar and sub-tropical jet streams from January 7th 11th at 12Z respectively, as simulated by the UW-NMS. Upper Left Synoptic setup for this orographic storm event.

38 33 b. Mid-Levels Large Scale Potential Vorticity As was shown in the synoptic schematic in figure 4, there were many undercutting disturbances steered by the jet streams aloft that enhanced the precipitation over the California coast. One of the best tools to diagnose these disturbances or short waves is to analyze the potential vorticity field. Potential vorticity is a very useful quantity mainly because it is conserved on isentropic surfaces. The following equation is the mathematical definition of isentropic potential vorticity. Equation 5. Mathematical definition of potential vorticity. P Isen. PV g Gravity ζ Relative Vorticity Θ f Coriolis Parameter d θ / dp Diff. change of Theta with Pressure Since the focus of this section is the mid-level dynamics, 700mb potential vorticity is used as a diagnostic tool. At this level various short waves should be evident. Shown in figure 5 is the 700mb potential vorticity, as simulated by the UW-NMS, beginning on January 7 th th at 12Z through January 11 at 12Z. Very evident is the large trough off the Pacific Northwest. High potential vorticity is associated with this trough due to the enhanced cyclonic flow, potentially showing the presence of a tropopause fold. The focus of this section is the various short waves outside the main cyclonic flow. On January 8 th, 9 th, and 10 th these short waves are evident by an increase of PV. It is evident that these waves continuously propagate towards the coast during this time. Also shown in figure 5 is the 700mb streamline flow. This is useful for qualitatively diagnosing regions of potential vorticity advection. From quasi-geostrophic theory, regions of positive vorticity advection are favored areas for rising motion. This